[RFC] standardise exit-codes

Writing this down, as I was thinking about this a while back if there's a way to standardise exit codes (also in light of https://github.com/moby/moby/pull/38689 ("Make client return "rich" errors")), and https://github.com/docker/cli/pull/1674 got me thinking again 😅

The CLI currently uses "chroot" conventions for exit codes, where some exit-codes (> 125) have a special meaning; https://www.tldp.org/LDP/abs/html/exitcodes.html, and exit codes < 125 indicate exit codes from the container's process itself
(see https://github.com/moby/moby/pull/14012)

https://www.tldp.org/LDP/abs/html/exitcodes.html mentions an existing conventions/standard for C/C++.

I was wondering if it would make sense to map (API) errors to those C/C++ standard (if possible); https://github.molgen.mpg.de/git-mirror/glibc/blob/0f1e40d49c1d4871fa9d7b7efdc7fca7f5ac01c0/misc/sysexits.h#L47-L89

Or perhaps there's another standard that makes sense?

For example (this is really a quick "blurb");

Start by reviewing the Docker CLI exit-code conventions described in the issue and the linked moby/moby and docker/cli pull requests. Compare the proposed sysexits.h mappings with the current behavior for container exits, API errors, and unavailable services. Done means the project has agreed on a standard mapping and a concrete implementation scope.